AI output has patterns everyone recognizes.

One file removes them permanently.

_

It's called anti-ai-writing-style.md

Claude runs it as a self-audit before showing you any output.

_

Here's what to put inside it:

1. Banned words.

"Leverage," "unlock," "game-changer," "seamlessly," "dive in," "delve."

Add every word that makes you cringe when you see it.

_

2. Banned sentence structures.

Rule of three. Staccato blunt lines. "Not just X but Y."

Any structural pattern that reads like a LinkedIn cliché.

_

3. Banned formatting.

Em dashes for drama. Bold for emphasis. Rhetorical questions.

Bullet points when a sentence would do the job better.

_

4. Tone markers.

Anything that sounds corporate, performed, or overly polished.

"In today's fast-paced world." "More than ever before." "It's worth noting."

_

Then open Cowork and paste this:

Interview me about every AI writing pattern I hate. Ask about banned words, sentence structures, formatting habits, and tone markers. Write an anti-ai-writing-style.md file from my answers. Be specific and exhaustive. Save it to my about-me folder.

Claude interviews you.

The more specific your answers, the better the file.

_

Then go to Settings > Cowork > Global Instructions and add:

Before sharing any output, run the anti-ai-writing-style .md audit and fix every violation before showing me anything.

Claude checks itself and you stop seeing the same patterns twice.

_

Find the full breakdown here: ruben.substack.com/p/it…

♻️ Restack this if you learned something.